Benchmarks Are Broken and We're All Pretending Otherwise
Every major lab ships a model, immediately publishes a chart where their model wins, and calls it progress. I'm done buying it.
Essays on AI systems, software engineering, and mathematical thinking. Static HTML, RSS-first, and SEO-generated.
Every major lab ships a model, immediately publishes a chart where their model wins, and calls it progress. I'm done buying it.
Strong static types aren't a preference — they're the difference between a system and a pile of guesses.
Most engineers who think they can't do math actually can. The real problem is staying with confusion long enough for understanding to form.